Comparing Short-Term Thunderstorm Forecasts from the Warn-on-Forecast System (WoFS) and the High-Resolution Rapid Refresh (HRRR)

Short-term, 0–6 hour, forecasts of thunderstorms from the experimental Warn-on-Forecast System (WoFS) are compared to forecasts from the operational High-Resolution Rapid Refresh (HRRR) for 95 cases between 2020 and 2023. Object-based and neighborhood verification of simulated composite reflectivity from the two systems against Multi-Radar Multi-Sensor (MRMS) observations is used to quantify differences in forecast quality for “next hour” prediction of thunderstorms. Specific errors are examined to estimate impacts of configuration differences between the two systems, including: Does WoFS probabilistic guidance improve over deterministic forecasts from the HRRR or WoFS members? Does more frequent data assimilation in WoFS improve thunderstorm forecast accuracy, particularly near convection initiation?

Results show that the primary advantage of WoFS over HRRR for short-term thunderstorm prediction is the ability to provide probabilistic guidance, which provides a significant improvement over deterministic forecasts from the HRRR or WoFS members at all lead times. Deterministically, the HRRR and WoFS members predict thunderstorms with similar accuracy for the first three hours of the forecast, after which HRRR provides significantly more accurate forecasts. Modest improvements are found in WoFS members that may be attributable to more frequent data assimilation, including reduced storm displacement errors and higher probability of detection in the hour prior to convection initiation. WoFS member predictions of thunderstorms are limited by excessive spurious convection relative to the HRRR at all lead times. Additionally, errors in the boundary conditions for the 900 km x 900 km WoFS domain contribute to strongly degraded forecast quality near the western and southern boundaries, and potentially to degraded forecast quality throughout the domain for lead times greater than three hours.